N.J. Ferrier, B.E. Klein, and L. Hubbard (USA)
ophthalmic image analysis, cataract, unsupervised learning
We present a method to classify images using intensity val ues of features that have been localized in slit lamp images of the lens. A multi-step algorithm is used to train a lin ear self-organizing map to determine the mapping from the image feature space to a scalar grading scale. Using a large set of images for unsupervised training, we evaluate our resulting function using a small set of images whose clas siļ¬cation has been provide by human experts. Results are presented showing that a high correlation is achieved.
Important Links:
Go Back
